Hi all,
Currently I am running into a rather puzzling bug. I have a simple input deck that causes an error in MPI_Allreduce if I run it with more than one core, and no errors if I run it with one core.
If I comment out a fix momentum, the error disappears. I am, however, fairly certain that these types of input used to work fine… Maybe it has to do with Arch Linux moving to OpenRTE or something…
Can anyone reproduce this crash?
$ mpirun -np 1 lmp -in bug.in -echo screen # No crash
$ mpirun -np 2 lmp -in bug.in -echo screen # Error in MPI_Allreduce
mpirun -np 1 lmp -in [nobug.in](http://nobug.in) -echo screen # No crash
mpirun -np 2 lmp -in nobug.in -echo screen # No crash
Thanks!
bug.in (508 Bytes)
nobug.in (509 Bytes)
resume.data (463 KB)
Hi all,
Currently I am running into a rather puzzling bug. I have a simple input
deck that causes an error in MPI_Allreduce if I run it with more than one
core, and no errors if I run it with one core.
If I comment out a fix momentum, the error disappears. I am, however, fairly
certain that these types of input used to work fine... Maybe it has to do
with Arch Linux moving to OpenRTE or something...
Can anyone reproduce this crash?
which LAMMPS version exactly is this with?
this same issue was reported recently and it should be addressed with
this commit from May 25th to LAMMPS-ICMS:
https://github.com/lammps/lammps/commit/4c5763fa8551ca70c1ca9fedcc78e3e46776af06
we're currently in the process of merging LAMMPS-ICMS to upstream
after a three week break, thus it should be in the official version
shortly.
axel.
Oh, right, my version is 14 May. I will update and check again, completely forgot about this.
Yup, this fixed it, thanks!